A5082228369- Remote-Sensing Image Classification
- Remote Sensing and Land Use
- Wireless Networks and Protocols
- Metallurgical Processes and Thermodynamics
- Advanced Image Fusion Techniques
- Land Use and Ecosystem Services
- Mobile Ad Hoc Networks
- Advanced Computational Techniques and Applications
- Image and Signal Denoising Methods
- Advanced Decision-Making Techniques
- Remote Sensing in Agriculture
- Advanced Image and Video Retrieval Techniques
- Smart Agriculture and AI
- Cooperative Communication and Network Coding
- Software Engineering Techniques and Practices
- Industrial Technology and Control Systems
- Software Reliability and Analysis Research
- Bayesian Modeling and Causal Inference
- Evaluation and Optimization Models
- Advanced Chemical Sensor Technologies
- Rough Sets and Fuzzy Logic
- Mathematical and Theoretical Epidemiology and Ecology Models
- Service-Oriented Architecture and Web Services
- Advanced Vision and Imaging
- Robotics and Sensor-Based Localization
Beijing Normal University
2022-2025
State Key Laboratory of Remote Sensing Science
2010-2024
Hebei University of Science and Technology
2013-2024
Harbin Institute of Technology
2006-2024
Mitchell Institute
2024
Texas A&M University
2024
Henan Academy of Sciences
2024
Wuhan University
2010-2024
Ocean University of China
2023
University of Science and Technology Beijing
2000-2023
In many computer vision applications, obtaining images of high resolution in both the spatial and spectral domains are equally important. However, due to hardware limitations, one can only expect acquire either or domains. This paper focuses on hyperspectral image super-resolution (HSI-SR), where a (HSI) with low (LR) but is fused multispectral (MSI) (HR) obtain HR HSI. Existing deep learning-based solutions all supervised that would need large training set availability HSI, which...
Linear spectral unmixing is the practice of decomposing mixed pixel into a linear combination constituent endmembers and estimated abundances. This paper focuses on unsupervised where are unknown priori. Conventional approaches use either geometrical- or statistical-based approaches. In this paper, we address challenges with deep learning models, in specific, autoencoder decoder serves as hidden layer output several recent attempts, part-based autoencoders have been designed to solve...
Domain adaptation techniques have been widely applied to the problem of cross-scene hyperspectral image (HSI) classification. Most existing methods use convolutional neural networks (CNNs) extract statistical features from data and often neglect potential topological structure information between different land cover classes. CNN-based approaches generally only model local spatial relationships samples, which largely limits their ability capture nonlocal relationship that would better...
Anomaly detection has been known to be a challenging problem due the uncertainty of anomaly and interference noise. In this paper, we focus on in hyperspectral images (HSIs) propose novel algorithm based spectral unmixing dictionary-based low-rank decomposition. The innovation is threefold. First, highly mixed nature pixels HSI data, instead using raw pixel directly for detection, proposed applies obtain abundance vectors uses these detection. We show that possess more distinctive features...
The key to hyperspectral anomaly detection is effectively distinguish anomalies from the background, especially in case that background complex and are weak. Hyperspectral imagery (HSI) as an image-spectrum merging cube data can be intrinsically represented a third-order tensor integrates spectral information spatial information. In this article, prior-based approximation (PTA) proposed for detection, which HSI decomposed into tensor. tensor, low-rank prior incorporated dimension by...
Hyperspectral target detection can be described as locating targets of interest within a hyperspectral image based on prior information targets. The complexity actual scenes limits the performance traditional statistical methods that rely model assumptions, and machine learning mapping functions with limited complexity. To address these problems, we propose Siamese transformer network for (STTD). contribution this article is threefold. First, novel method constructing training samples using...
Pansharpening is to fuse a multispectral image (MSI) of low-spatial-resolution (LR) but rich spectral characteristics with panchromatic (PAN) high-spatial-resolution (HR) poor characteristics. Traditional methods usually inject the extracted high-frequency details from PAN into up-sampled MSI. Recent deep learning endeavors are mostly supervised assuming HR MSI available, which unrealistic especially for satellite images. Nonetheless, these could not fully exploit in Due wide existence mixed...
The purpose of hyperspectral anomaly detection is to distinguish abnormal objects from the surrounding background. In actual scenes, however, complexity ground objects, high-dimensionality data and non-linear correlation bands have high requirements for generalizability, feature extraction ability nonlinear expression algorithms. order address above problems, we propose a transferable network with Siamese architecture image (TSN-HAD). contribution TSN-HAD three-fold. First, problem through...
Abstract Most unsupervised or semisupervised hyperspectral anomaly detection (HAD) methods train background reconstruction models in the original spectral domain. However, due to noise and spatial resolution limitations, there may be a lack of discrimination between backgrounds anomalies. This makes it easy for autoencoder capture low‐level features shared two, thereby increasing difficulty separating anomalies from backgrounds, which runs counter purpose HAD. To this end, authors map...
Pansharpening refers to the fusion of a high spatial resolution panchromatic image with spectral multispectral or hyperspectral images (MSI HSI) yield data in both and domains. It has been widely adopted as primary preprocessing step for numerous applications. In this paper, we perform literature survey various pansharpening algorithms including most advanced deep learning approaches images. We further evaluate effect difference on anomaly detection. Synthetic are generated performance...
Target detection aims to locate targets of interest within a specific scene. The traditional model-driven detectors based on signal processing have proved be very effective. However, the performance such methods relies heavily model assumption, which is limited by discrepancy with real hyperspectral images (HSIs) data. In this article, target method through tree-structured encoding (TD-TSE) for HSIs proposed. Instead modeling and background extract valid features, we construct binary tree...
Africa carries a disproportionately high share of the global malaria burden, accounting for 94% cases and deaths worldwide in 2019. It is also politically unstable region most vulnerable continent to climate change recent decades. Knowledge about modifying impacts violent conflict on climate–malaria relationships remains limited. Here, we quantify associations between conflict, variability, risk sub-Saharan using health surveys from 128,326 individuals, historical data, 17,429 recorded...
Anomaly detection has been known to be a challenging, ill-posed problem due the uncertainty of anomaly and interference noise. In this paper, we propose novel low rank algorithm in hyperspectral images (HSI), where three components are involved. First, highly mixed nature pixels HSI, instead using raw pixel directly for detection, proposed applies spectral unmixing algorithms obtain abundance vectors uses these detection. Second, better classification, dictionary is built based on mean-shift...
Convolutional neural networks (CNNs) have been widely applied in hyperspectral imagery (HSI) classification. However, their classification performance might be limited by the scarcity of labeled data to used for training and validation. In this paper, we propose a novel lightweight shuffled group convolutional network (abbreviated as SG-CNN) achieve efficient with dataset HSI SG-CNN consists SG conv units that employ conventional atrous convolution different groups, followed channel shuffle...
Given the prior information of target, hyperspectral target detection focuses on exploiting spectral differences to separate objects interest from background, which can be treated as retrieval (IR) task in machine learning (ML). Most traditional methods work original feature space and rely heavily specific assumptions, cannot guarantee effective extraction features for background images (HSIs). Mass estimation (ME) is a base modeling mechanism that has been proven effectively solve problems...
Hyperspectral image (HSI) classification is one of the most active research topics and has achieved promising results boosted by recent development deep learning. However, state-of-the-art approaches tend to perform poorly when training testing images are on different domains, e.g., source domain target domain, respectively, due spectral variability caused acquisition conditions. Transfer learning-based methods address this problem pre-training in fine-tuning domain. Nonetheless, a...
This paper reviewed the various factors affecting removal of PFASs from water by nanofiltration and reverse osmosis.
The elderly face elevated mortality risk due to rising temperature. Previous assessments of temperature-related mortality, however, lack a comprehensive analysis distinct impacts temperature change across different timescales and characteristics. Using longitudinal survey 27,233 Chinese citizens from 2005 2018, we establish connections between temperatures, variability, extreme heat with increased risk, assessed through four annual metrics that combine humidity. intensity prolonged duration...
With the rapid development of urbanization in China, a large number small towns have emerged. These face problems such as limited resources, backward industries, and aging infrastructure. How to achieve sustainable is crucial for China implement its goals. Based on status existing research towns, this paper develops performance evaluation index system encompassing five key dimensions, employs system, along with maturity assessment, appraise towns. Applying process Tianhuangping Town, results...
Scene classification is a fundamental task for numeral remote sensing applications, which aims to assign semantic labels image patches. Although deep neural networks (DNN) demonstrated unique strength in scene classification, their performances are still limited due the lack of training samples field. Recent studies show that performance can be improved by taking advantage knowledge transferred from models pre-trained on RGB images. However, modalities differences between input images hinder...
China contributed nearly one-fifth of the world maize production over past few years. Mapping distributions cropland in is crucial to ensure global food security. Nonetheless, 10 m maps are still unavailable, restricting promotion sustainable agriculture. In this paper, we collect numerous samples produce annual 10-m from 2017 2021 with a machine learning based classification framework. To overcome temporal variations plants, proposed framework takes Sentinel-2 sequence images as input and...